<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Luka's blog</title><link>https://www.sugarz.me/</link><description>Recent content on Luka's blog</description><generator>Hugo -- gohugo.io</generator><language>en-gb</language><lastBuildDate>Fri, 24 Apr 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://www.sugarz.me/index.xml" rel="self" type="application/rss+xml"/><item><title>Porting V-JEPA 2.1 to MLX</title><link>https://www.sugarz.me/posts/porting_vjepa_to_mlx/</link><pubDate>Fri, 24 Apr 2026 00:00:00 +0000</pubDate><guid>https://www.sugarz.me/posts/porting_vjepa_to_mlx/</guid><description>I ported the V-JEPA 2.1 encoder to MLX, verified output parity against the upstream PyTorch checkpoints, and benchmarked the result on Apple silicon.</description></item><item><title>Can Cyrillic survive the age of AI?</title><link>https://www.sugarz.me/posts/cyrillic_llm/</link><pubDate>Mon, 07 Apr 2025 18:24:19 +0200</pubDate><guid>https://www.sugarz.me/posts/cyrillic_llm/</guid><description>&lt;p&gt;As we become increasingly dependent on AI, an important question arises: Will this dependency help smaller languages and alphabets survive, or will it hasten their path toward oblivion? Will AGI ever speak Serbian?&lt;/p&gt;
&lt;p&gt;As Serbian language can be written using two different alphabets, which one will the LLMs think in?
It&amp;rsquo;s a tough question, and in this blog post I&amp;rsquo;ll try to explore LLM&amp;rsquo;s knowledge of both alphabets and the implications it has on current writing patterns and data availability.&lt;/p&gt;</description></item><item><title>About</title><link>https://www.sugarz.me/pages/about/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://www.sugarz.me/pages/about/</guid><description>&lt;p&gt;My name is Luka Šećerović, and I&amp;rsquo;m an ML engineer. Here I write about my projects. The blog is still early, but there&amp;rsquo;s a lot more coming.&lt;/p&gt;
&lt;p&gt;TL;DR:
I&amp;rsquo;ve spent 7 professional years building software and ML systems at Microsoft and a couple of startups, and most of that work is closed source. I always wanted to do more open source work and share it with the world, so this blog is where I&amp;rsquo;ll do more of that and write about what I learn along the way.&lt;/p&gt;</description></item></channel></rss>